Hearthloom

Local by design

The story
Hearthloom home intelligence appliance

Your own private AI

Choose how much AI
your home can run.

Every Hearthloom replaces the everyday need for hosted AI chat, runs coding and agent work locally, creates feeds with your algorithm, and can be upgraded. Choose by how much needs to happen at once.

Three household scales

Same principles.
More room to think.

01

Target $899

Core

1–4 people
1 intensive session

For an individual, couple, or family that wants fast private chat, serious coding, and a primary agent while Hearthloom continues its household work.

  • Qwen3.8 27B · four-bit
  • Hearthloom mobile app
  • Private chat, coding, agents, feeds
  • 1TB local storage

Typical dayA sustained chat or coding session, a personal feed, household monitoring, and quick requests from another person.

02

Target $1,499

Family

4–8 people
2–4 intensive sessions

For busy households where several people may need sustained help while Hearthloom’s background agents continue working.

  • More CUDA memory
  • Longer working context
  • More local storage
  • Several private assistants

Typical dayCoding, document analysis, homework help, personal agents, feeds, and home awareness at the same time.

03

Target $2,699

Estate

8–20+ people
6–12 intensive sessions

For large households, staff, guests, multiple buildings, and continuously operating property intelligence.

  • Expandable acceleration
  • Property and maintenance agents
  • High-volume sensor analysis
  • Family, staff, and guest profiles

Typical daySeveral people chatting and creating while research, property, security, energy, and maintenance agents work concurrently.

Intensive session means sustained conversation, document or image analysis, research, or agent work. Simple home commands and background awareness continue alongside these sessions.

Included in every Hearthloom

The complete experience.
Not just the box.

Mobile app

Chat instead of ChatGPT

Talk to Hearthloom from your phone while the conversation and memory remain on the machine at home by default.

Work

Coding and agents

Write and debug code, analyze files, and give persistent private agents jobs to complete.

Discovery

Your own feeds

Choose sources, ranking rules, and goals instead of accepting an engagement company’s algorithm.

Intelligence

Qwen3.8 27B

A powerful open model, locally optimized at four-bit precision and accelerated by NVIDIA CUDA.

AAA AI.
At the cost of electricity.

Local conversations, coding, feed-building, and agent work do not create a per-token cloud bill. No OpenAI, Google, Anthropic, or Grok account is required. Optional outside services remain your choice.

Hearthloom installed with a home stereo and entertainment system

Made to be kept

Change the machine.
Keep the home.

Models, compute, encrypted storage, and connections can advance independently. The household’s identity, permissions, and chosen memories do not have to start over.

ModelsInstall, compare, or roll back open models.

ComputeReplace the GPU or compute sled as local AI advances.

StorageExpand and back up encrypted household memory.

ConnectionsAdopt new device standards through serviceable modules.

Secure is visible

It tells you
what it knows.

The display reports meaningful household state instead of demanding attention.

HOME SECUREEXTERNAL NETWORK OFF

EVERYTHING CALMNO ACTION NEEDED

Close view of Hearthloom with its household status display illuminated

The intelligent future is yours

Own your AI.

Start with the capacity your household needs. Chat, code, run agents, and shape your own feeds on intelligence that remains yours.

Return to the Hearthloom story